ID_Repo_Choice

ID_Repo_Choice is the configuration node of the ID_Animator pack - the one that decides which SD 1.5 base, which dreambooth checkpoint, and which AnimateDiff motion module the ID_Animator node will use. It exists because ID_Animator's repo_id input is a force-input STRING that carries three model paths at once. This node packages them into that string for you, so you never hand-type a comma-delimited mess.

It's not flashy. It's the node that prevents the "I installed everything and it still won't run" moment, which in this pack is a real moment.

What it outputs

A single repo_id STRING that's actually three comma-joined values: the diffusers repo_id (an HF repo or a local folder), the dreambooth checkpoint path, and the motion module path. Wire it straight into ID_Animator's repo_id input - the two were built to talk to each other.

The sneaky good part: if you leave motion_model on "none", the node auto-downloads mm_sd_v15_v2.ckpt from guoyww/animatediff into models/animatediff_models. Same trick ID_Animator uses for the domain adapter. That's one less thing for you to hunt down.

The inputs you'll actually touch

• local_model_path - dropdown of the diffusers folders it finds under ComfyUI/models/diffusers; stays "none" until you've put an SD 1.5 diffusers folder there.
• repo_id - string, default runwayml/stable-diffusion-v1-5. This is what gets used if you don't pick a local folder. Keep the owner/name format.
• ckpt_name - dropdown of your ComfyUI/models/checkpoints. Pick a dreambooth-style SD 1.5 checkpoint (realisticVision or similar) here to control style; it's optional but it's how you get output that doesn't look like raw vanilla 1.5.
• motion_model - dropdown of files in models/animatediff_models, plus "none" for the auto-download.

Output: repo_id (STRING) → ID_Animator's repo_id input.

Install

Same pack, same steps as ID_Animator: ComfyUI-Manager search "ComfyUI_ID_Animator", or

cd ComfyUI/custom_nodes
git clone https://github.com/smthemex/ComfyUI_ID_Animator.git

then restart. And same warning: the pack's requirements.txt pins old torch/xformers/transformers versions, so don't let the installer clobber your main environment - install the missing pieces individually (insightface, onnxruntime, einops, imageio, omegaconf, diffusers 0.28+).

Gotchas

• Type a repo_id without a / and the node raises "Incorrect repo_id format" - HF repos are owner/name. Empty repo_id plus local_model_path = "none" raises "you need fill repo_id or download model in diffusers directory". One caveat worth knowing: those errors are raised as bare strings in the code, so in modern Python they surface as a TypeError crash with the message buried in the traceback. The text is still there - it just isn't a clean red banner.
• The local_model_path dropdown is built at load time, so if you drop a new diffusers folder in, restart ComfyUI before it shows up.
• Every choice here only matters because AnimateDiff is an SD 1.5 technology - the motion module won't transfer to SDXL or a dedicated video model, so point it at 1.5 checkpoints and nothing else.